How Real Fake News on tinyBots.net Works
Updated
Real Fake News is a tinyBots.net project that turns a headline you supply into a full news article and a matching image. You provide the headline; the service sends it to OpenAI's ChatGPT API to generate the article text, then attempts to create a relevant image through the AI Horde, an open-sourced Stable Diffusion distributed cluster. The result is an automatically written, illustrated news story — described on the site as "what will happen when robots take over The Onion."
The Two-Step Generation Pipeline
The project chains two separate AI services, each handling one part of the output.
Step 1: Article text via ChatGPT
- Input: a news headline you provide.
- Action: the service makes a request to OpenAI's ChatGPT API.
- Output: a corresponding news article written to match that headline.
The headline is the only creative input described on the site. Everything in the body text is generated by ChatGPT in response to it, so the tone and content of the article follow whatever framing the headline sets.
Step 2: Illustration via AI Horde
- Action: after the article is generated, the service attempts to use the AI Horde to produce a relevant image.
- Output: an image intended to accompany the generated article.
Note the wording on the site: it "attempts to" generate an image. The AI Horde is a distributed cluster of volunteer-run Stable Diffusion workers, so image generation depends on that shared capacity rather than a dedicated, guaranteed resource. The article text and the image are produced by different systems, which means the illustration is generated to be relevant to the story rather than being a literal, verified depiction of it.
Why the Two Services Are Split
The same AI Horde cluster also powers ArtBot, another tinyBots.net project for generative AI art. Reusing it for Real Fake News keeps the image step on free, open-sourced infrastructure while the text step relies on OpenAI's API. That split is the core mechanism: a commercial text model for the writing, a distributed open-source image model for the picture.
What to Expect in Practice
- You control the premise, not the details. The headline steers the article; the specifics are invented by the model.
- Text and image are independent. A coherent article and a thematically related image can still diverge in specifics.
- Image generation may not always complete. Because it depends on the AI Horde's shared cluster, treat the illustration as best-effort rather than guaranteed.
- The output reads as satire. The site frames the whole thing as automated Onion-style news, so the intended use is humorous or experimental rather than factual reporting.
Related tinyBots.net Projects
| Project | What it does | AI dependency |
|---|---|---|
| Real Fake News | Headline → article + image | ChatGPT API + AI Horde |
| ArtBot | Generates AI art from prompts | AI Horde (free) |
| Mr. RossBot | Tweets a Bob Ross–style landscape hourly | Stable Diffusion |
| Just Triangles | Procedurally generates triangle patterns | None (uses colormind.io palettes) |
| Random Sandwiches | Generated "book" of 1,449.5 sandwich recipes | None (NaNoGenMo 2017) |
If your goal is generating a full illustrated article from a single headline, Real Fake News is the relevant tool; if you only want images, ArtBot covers that separately.